COVID-19 Rate Tracker

The graphs on the left are rolling 7 day sums of past data reported by the WHO and The Covid Tracking Project.
We use the closest thing we can find to a single source of raw data. This is as close to 'real' as we could find.
They are normalized by the reporting region's population in millions. This provides people all over the world with a common value for goal setting and reporting. It also allows the data to be compared.
A data point is computed for each WHO reporting region for each day.
The data point is the sum of the past 7, 14 or 21 days (including the current day).
7, 14 and 21 were chosen because each day of the week has different reporting behavior. People don't report the same way on Sundays and Mondays. This way, data for each day of the week is equally represented in the sample.
On the right are Normalized Cases Rate Slope (NCRS) or Normalized Death Rate Slope (NDRS) graphs. These are computed using a least squares linear fit of the normalized weekly rate data.
The data point is the slope of the least squares linear fit for the smoothed data taken from the preceding graph. Not smoothing the data produces unintelligible noise.
The number of samples in the slope calculation are equal to the smoothing sample size (7, 14 or 21).
This smoothing process allows untrained readers to see the same trends that experienced data analysts can see by looking at the raw data.

For details of the computations click here. By Dark Sky Innovative Solutions. Add the code to your site from gitHub.com US State data provided by The Covid Tracking Project

Smoothing days:7 14 21 28

Loading WHO data. Please wait.

Loading USA states data. Please wait.

Sorted By Closure Until Zero New Cases
Days To Zero Is An Indicator Only4

Data is loading, please wait. This normally takes less than one minute and only happens once each time the page loads.

7 Day New Cases per / million
Days To Zero Is An Indicator Only

Data is loading, please wait. This normally takes less than one minute and only happens once each time the page loads.

Sorted By Closure Until Zero New Cases
Days To Zero Is An Indicator Only4

Data is loading, please wait. This normally takes less than one minute and only happens once each time the page loads.

7 Day New Cases / million
Days To Zero Is An Indicator Only4

Data is loading, please wait. This normally takes less than one minute and only happens once each time the page loads.

COVID-19 Daily Cost of Treatment1
Current cost estimates are about $7,000/day per patient.2

Data is loading, please wait. This normally takes less than one minute and only happens once each time the page loads.

Details

This is fundamental, high school math, historical rate and trend tracking. There are NO PREDICTIONS of future results. If you want predictions, see an actuary or bookie.

However, if you track the slope of case rates curve for your region, your data collection is of 'good quality' and you keep your slope of rates negative for 'long enough,' it is reasonable to expect that your region will enter a maintenance state such as the Republic of Korea or Japan have. Even if your region experienced extremely high rates (such as Andorra, Belgium, Italy and Singapore), your region can recover as they have - if you can keep your slope of case rates negative.

How to keep the slope of case rates negative is up to local citizens, health officials and governments. Dark Sky Innovative solutions does not attempt to give advice on this topic.

From observation of these historical trends, the challenge is not deciding when to lock down, but when to open and when to lock down again. The slope of rates graph is a good indicator of when you should or should not reopen. The curve uses the latest data available and the average lags by about 1 week. But if there is a statistically large increase in rates, the curve will show that immediately.

It is my opinion, that in a crisis with so many unknowns, excess caution and responsiveness are more crucial to success than statistical or mathematical rigor. Severely affected regions simply don't have the time to wait for thorough mathematical analyses.

"If your sample is large enough and selected properly, it will represent the whole well enough for most purposes. If it is not, it may be far less accurate than an intelligent guess and have nothing to recommend it but a spurious air of scientific precision." - How to Lie With Statistics by Darrell Huff copyright 1954.

The above quote explains why the actuaries and bookies aren't able to make predictions. There's not enough history yet.

The methods I used to construct these curves have served me well for 40 years and have been crucial to successful projects which were being overwhelmed by noisy, dare I say crappy, data.

Paralysis by analysis can kill when it applies to COVID-19

Bob Tipton - Founder Dark Sky Innovative Solutions

Hospital data is processed differently and of very low reliability, not provided for countries nor by several states due to changes in reporting ordered by the US President. In cases where data has been cut off we report the last known value. Since hospitals were near capacity at the time of the cutoff, this number should be reasonably accurate. Cost data is computed by taking number of hospital patients listed as COVID-19 and multiplying by $7,000, an estimated figure form an Atlantic Magazine report.
One obvious discrepancy is that New York state has continuing deaths, ~800 people in the hospital but zero admitted to the hospital for several weeks.
Not controlling COVID-19 has a cost also. Reporting costs while people are dying is usually considered morbid. Since one of the reasons our leaders use to delay controlling COVID is cost, and the best way to reduce costs and save lives is to control it, reporting the cost may motivate our leaders to save lives.
Current estimates are about $6,000 to $8,000 per patient per day. This graph shows the current number of patients reported in the hospital for COVID-19 mulitplied by cost per day for treatment.
Early in the Pandemic, the WHO reported that COVID-19 had a low death rate when properly treated. They made no statement about the cost of that treatment to the community.
Deniers claim that the USA has a very low fatality rate relative to other countries.
First, we are in the top 10 or 20 for Confirmed Fatality Rate by rank order. Our arithmetic mean is near the world average.
Second, the US is paying a price to keep patients alive.
On 20 July 2020, the NCRS slope calculation was changed to always use the last 7 days of smoothed data. Previously the slope sample size was equal to the smoothing window size. I concluded that 14 or 21 day sums were sufficient for smoothing and the slope data would respond faster and reflect current slope better with a fixed 7 day sample size.
On 3 August 2020, days until zero were added if the value was less than one year. This value is displayed with an orange background to bring attention to the fact that IT IS NOT A PREDICTION! If this rate could be maintained, the region would reach zero new cases/week in this number of days. This is virtually impossible and is only provided as a guide!
If a region uses this guide to control their new case rate, it will inform them of their progress in reaching and maintaining zero cases.
The quality of the value is poor at best. It is based on data reported to WHO and The Covid Tracking Project. To have greater reliability, a region must test more thoroughly and report data accurately to these data centers. The Covid Tracking Project gets much of it's data from USA CDC.
Dark Sky Innovative Solutions provides this data as a public service and states that it's quality is doubtful. However, we firmly believe it is the best indicator available for tracking containment and control of COVID-19.

On 27 July 2020, the list sorted by NCRS was replace with a list sorted by 'closing on zero'. If the NCRS is negative, the intercept of the current slope is computed. If a region has a new case rate of 100 and an NCRS = -10 ideally they would reach zero new cases in 10 days. Since this number is totally unpredictable in the real world, we do not display it - but we do sort by it.
If the NCRS is positive (there is no closure yet) we sort by that instead.
Regions which are closing on zero are green.
Regions moving away from zero are red.
On 25 March the data source for US states was changed from Covid Tracking Project to US CDC. Hospitalization data is not available at this time.